"Faking" a sub-table of results from 3 tables, in SQL - mysql

Question

If I have a table of cases:
CASE_NUMBER | CASE_ID | STATUS | SUBJECT |
----------------------------------------------------------------
3108 | 123456 | Closed_Billable | Something Interesting
3109 | 325124 | Closed_Billable | Broken printer
3110 | 432432 | Open_Assigned | Email not working
And a table of calls:
PARENT_ID | STATUS | DUR(H) | DUR(M) | SUBJECT
---------------------------------------------------------------
123456 | Held | 1 | 30 | Initial discussion
123456 | Cancelled | 0 | 0 | Walk user through
123456 | Held | 0 | 45 | Remote debug session
325124 | Held | 1 | 0 | Consultation
325124 | Held | 1 | 15 | Needs assessment
432432 | Held | 1 | 30 | Support call
And a table of meetings:
PARENT_ID | STATUS | DUR(H) | DUR(M) | SUBJECT
-------------------------------------------------------
123456 | Held | 3 | 15 | On-site work
325124 | Held | 2 | 0 | Un-jam printer
432432 | Held | 1 | 0 | Reconnect network
How do I do a select with these parameters (this is not working code, obviously):
SELECT cases.case_number, cases.subject, calls.subject, meetings.subject
WHERE cases.status="Closed_Billable" AND (calls.status="Held" OR meetings.status="Held)
LEFT JOIN cases
ON cases.case_id = calls.parent_id
LEFT JOIN cases
ON cases.case_id = meetings.parent_id
and end up with a "faked" nested table like:
CASE_NUMBER | CASE SUBJECT | # CALLS | # MEETINGS | CALL SUBJECT | MEETING SUBJECT | DURATION (H) | DURATION (M) | TOTAL
-----------------------------------------------------------------------------------------------------------------------------------------
3108 | Something Interesting | 2 | 1 | | | | | 5.5H
| | | | Initial Discussion | | 1 | 30 |
| | | | Remote Debug Session | | 0 | 45 |
| | | | | On-site work | 3 | 15 |
3109 | Broken printer | 2 | 1 | | | | | 4.25H
| | | | Consultation | | 1 | 0 |
| | | | Needs assessment | | 1 | 15 |
| | | | | Un-jam printer | 2 | 0 |
I've tried joins and subqueries the best I can figure out, but I get repeated entries - for example, each Meeting in a Case will show say 3 times, once for each Call in that case.
I'm stumped! Obviously there's other fields I'm pulling here, and doing COUNTs of Calls and Meetings, and SUMs of their durations, but I'd be happy just to show a table/sub-table like this.
Is it even possible?
Thanks,
David.

Answer 1

Assembling a query result in the exact format you want is .. somewhat of a pain. It can be done, but presentation stuff like that is best left to the application.
That said, this will do what you want:
select case when case_id > floor(case_id) then ''
else case_number
end case_number,
coalesce(q1.c, '') calls,
coalesce(q2.c, '') meetings,
coalesce(calls.subject, '') `call subject`,
coalesce(meetings.subject, '') `meeting subject`,
case when calls.subject is not null then calls.dhour
when meetings.subject is not null then meetings.dhour
else ''
end dhour,
case when calls.subject is not null then calls.dmin
when meetings.subject is not null then meetings.dmin
else ''
end dhour,
coalesce(q3.total, '') total
from
(
select case_number, case_id
from cases where status = 'Closed_Billable'
union select case_number, concat(case_id, '.1')
from cases where status = 'Closed_Billable'
union select case_number, concat(case_id, '.2')
from cases where status = 'Closed_Billable'
) main
left join
(select parent_id, count(*) c
from calls
where status != 'Cancelled'
group by parent_id ) q1
on q1.parent_id = case_id
left join
(select parent_id, count(*) c
from meetings
group by parent_id) q2
on q2.parent_id = case_id
left join
(select parent_id, sum(dhour + m) total
from
(select parent_id, dhour, dmin / 60 m
from calls
where status != 'Cancelled'
union all
select parent_id, dhour, dmin / 60 m
from meetings
) qq
group by parent_id
) q3
on q3.parent_id = case_id
left join calls
on concat(calls.parent_id, '.1') = main.case_id
left join meetings
on concat(meetings.parent_id, '.2') = main.case_id
order by case_id asc
Note, i've renamed your duration fields because i dislike the parenthesis in them.
We have to mangle the case_id a little bit inside the query in order to be able to get you your blank rows / fields - those are what makes the query cumbersome
There's a demo here: http://sqlfiddle.com/#!9/d59d4/21
edited code to work with different schema in comment fiddle
select case when case_id > floor(case_id) then ''
else case_number
end case_number,
coalesce(q1.c, '') calls,
coalesce(q2.c, '') meetings,
coalesce(calls.name, '') `call subject`,
coalesce(meetings.name, '') `meeting subject`,
case when calls.name is not null then calls.duration_hours
when meetings.name is not null then meetings.duration_hours
else ''
end duration_hours,
case when calls.name is not null then calls.duration_minutes
when meetings.name is not null then meetings.duration_minutes
else ''
end duration_hours,
coalesce(q3.total, '') total
from
(
select case_number, id as case_id
from cases where status = 'Closed_Billable'
union select case_number, concat(id, '.1') as case_id
from cases where status = 'Closed_Billable'
union select case_number, concat(id, '.2') as case_id
from cases where status = 'Closed_Billable'
) main
left join
(select parent_id, count(*) c
from calls
where status != 'Cancelled'
group by parent_id ) q1
on q1.parent_id = case_id
left join
(select parent_id, count(*) c
from meetings
group by parent_id) q2
on q2.parent_id = case_id
left join
(select parent_id, sum(duration_hours + m) total
from
(select parent_id, duration_hours, duration_minutes / 60 m
from calls
where status != 'Cancelled'
union all
select parent_id, duration_hours, duration_minutes / 60 m
from meetings
) qq
group by parent_id
) q3
on q3.parent_id = case_id
left join calls
on concat(calls.parent_id, '.1') = main.case_id
left join meetings
on concat(meetings.parent_id, '.2') = main.case_id
order by case_id asc

Answer 2

You can't really get final results like that without some seriously ugly "wrapper" queries, of this sort:
SET #prevCaseNum := 'blahdyblahnowaythisshouldmatchanything';
SET #prevCaseSub := 'seeabovetonotmatchanything';
SELECT IF(#prevCaseNum = CASE_NUMBER, '', CASE_NUMBER) AS CASE_NUMBER
, IF(#prevCaseNum = CASE_NUMBER AND #prevCaseSubject = CASE_SUBJECT, '', CASE_SUBJECT) AS CASE_SUBJECT
, etc.....
, #prevCaseNum := CASE_NUMBER AS prevCaseNum
, #prevCaseSubject = CASE_SUBJECT AS prevCaseSub
, etc....
FROM ( [the real query] ORDER BY CASE_NUMBER, etc....) AS trq
;
And then wrap all that with another select to strip the prevCase fields.
And even this still won't give you the blanks you want on the "upper right".